reward scoring model enum

source/api_definitions.py

@@ -29,6 +29,7 @@ from model_types import (
     InstructModel,
     Message,
     PretrainedModel,
+    RewardModel,
     SamplingParams,
     ShieldConfig,
     StopReason,
@@ -245,9 +246,7 @@ class RewardScoringRequest:
     """Request to score a reward function. A list of prompts and a list of responses per prompt."""
 
     prompt_generations: List[KPromptGenerations]
-
+    model: RewardModel
-    # TODO(ragho): create a RewardModel enum tye
-    model: str
 
 
 @json_schema_type

source/model_types.py

@@ -137,3 +137,6 @@ class PretrainedModel(Enum):
 class InstructModel(Enum):
     llama3_8b_chat = "llama3_8b_chat"
     llama3_70b_chat = "llama3_70b_chat"
+
+class RewardModel(Enum):
+    llama3_405b_reward = "llama3_405b_reward"

source/openapi.html

@@ -1976,7 +1976,10 @@
                         }
                     },
                     "model": {
-                        "type": "string"
+                        "type": "string",
+                        "enum": [
+                            "llama3_405b_reward"
+                        ]
                     }
                 },
                 "additionalProperties": false,
@@ -2312,23 +2315,23 @@
         }
     ],
     "tags": [
+        {
+            "name": "RewardScoring"
+        },
         {
             "name": "Inference"
         },
-        {
-            "name": "SyntheticDataGeneration"
-        },
-        {
-            "name": "Datasets"
-        },
         {
             "name": "AgenticSystem"
         },
         {
-            "name": "Finetuning"
+            "name": "Datasets"
         },
         {
-            "name": "RewardScoring"
+            "name": "SyntheticDataGeneration"
+        },
+        {
+            "name": "Finetuning"
         },
         {
             "name": "ShieldConfig",

source/openapi.yaml

@@ -1039,6 +1039,8 @@ components:
       additionalProperties: false
       properties:
         model:
+          enum:
+          - llama3_405b_reward
           type: string
         prompt_generations:
           items:
@@ -1412,12 +1414,12 @@ security:
 servers:
 - url: http://llama.meta.com
 tags:
-- name: Inference
-- name: SyntheticDataGeneration
-- name: Datasets
-- name: AgenticSystem
-- name: Finetuning
 - name: RewardScoring
+- name: Inference
+- name: AgenticSystem
+- name: Datasets
+- name: SyntheticDataGeneration
+- name: Finetuning
 - description: <SchemaDefinition schemaRef="#/components/schemas/ShieldConfig" />
   name: ShieldConfig
 - description: <SchemaDefinition schemaRef="#/components/schemas/AgenticSystemCreateRequest"